风格还是内容？通过控制内容重叠评估风格分类器
Style or Content? Evaluating Style Classifiers with Controlled Content Overlap
AI总结：
提出控制内容重叠的评估方法，通过并行圣经翻译构建参数α，发现低重叠模型依赖内容线索，高重叠模型更鲁棒，为分离风格学习与内容捷径提供诊断。
摘要：
风格分类器可以利用自然收集数据中与风格标签相关的内容线索，但我们缺乏系统的方法来衡量这种依赖。我们通过基于并行圣经翻译构建的控制内容重叠设置来研究这个问题。具体来说，我们将重叠参数α定义为内容身份与风格标签之间互信息的归一化残差，从而衡量风格类别之间共享内容的程度：从无共享内容（α=0）到完全共享内容（α=1）。基于RoBERTa分类器的交叉重叠评估表明，当内容线索被移除时，低重叠模型性能下降，而高重叠模型迁移更鲁棒。跨风格内容检索探针进一步表明，随着α增加，内容变得难以恢复，训练动态显示这种移除是逐渐发生的。这些结果表明，控制重叠为分离风格学习与内容捷径提供了一个简单的诊断方法。
英文摘要：
Style classifiers can use content cues that correlate with style labels in naturally collected data, yet we lack a systematic way to measure this reliance. We study this problem with a controlled content overlap setup built on parallel Bible translations. Specifically, we define the overlap parameter $α$ as the normalized residual of mutual information between content identity and style label, so that it measures how much content is shared across style classes: from no shared content ($α=0$) to fully shared content ($α=1$). Cross-overlap evaluation of RoBERTa-based classifiers shows that low-overlap models degrade when content cues are removed, while high-overlap models transfer more robustly. A cross-style content retrieval probe further shows that content becomes less recoverable as $α$ increases, with training dynamics showing this removal occurs gradually. Together, these results suggest that controlled overlap provides a simple diagnostic for separating style learning from content shortcuts.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
